nace-ai
Nace.AI: Drex v1.5
nace-ai/drex-v1.5
Our take
AI-assisted analysis based on the sources below. Suitability is an editorial assessment; test it on your own workload.
Drex v1.5 is a roughly 9-billion-parameter decision model from Nace.AI, built on a Qwen3.5-style backbone distilled from XiaomiMiMo/MiMo-V2.6-Distill-Qwen-9B. Rather than generating text, it reads a state — text or JSON — plus typed questions (choice, yes/no, or ordinal score) and returns a calibrated probability for each option in a single forward pass per question. It exposes the System One /v1/systemone API, supports contexts up to 131,072 tokens, and can be self-hosted via Python/CUDA, llama.cpp, or an Ollama fork. Vendor-reported results place it at the top of the public Decision Index among sub-10B models, within that board's tie band.[1][2]
Strengths
- Structured decision output: per the model card, it scores choice, yes/no (noul), and ordinal-score questions with a calibrated probability per option in one forward pass, with no text generation.[1][2]
- Long context: 16,384 tokens by default, configurable up to 131,072; the vendor reports accuracy on long documents up to 128k tokens.[1][2]
- Flexible serving: one System One wire format across the bundled Python server, a llama.cpp branch (GGUF bf16/Q8_0), and a Nace Ollama fork; clients written for Kev or Jev are said to work unchanged.[2]
Limitations
- It is a scorer, not a generator: it only answers typed questions over a supplied state, and sampling parameters such as temperature do not apply.[2]
- Performance figures are self-reported: Decision Index, JevBench, long-context and game-arena numbers come from Nace.AI's own runs, and this pack contains catalog-plus-vendor-card evidence with no independent evaluation — validate on your workload.[1][2]
- Self-hosting has real requirements: about 18 GB of bf16 weights with a CUDA GPU for the Python path, or Nace forks of llama.cpp/Ollama for GGUF; released under the nonstandard Open RAIL-M license, and knowledge cutoff is undocumented.[1][2]
Where to use it
- Plausible fit for routing support tickets or triaging inbound requests into teams or categories, since the card's own example scores a billing-versus-shipping choice.[2]
- Worth evaluating for policy-style yes/no gates, such as detecting refund intent, where a calibrated probability is more useful than generated text.[2]
- Candidate for scoring or rating long documents against ordinal rubrics given the reported long-context accuracy, but confirm quality on your own data first.[2]
Sources
- [1] OpenRouter catalog · Nace.AI: Drex v1.5Sources checked: Nov 13, 2026
- [2] Model documentation · Nace.AI: Drex v1.5Sources checked: Nov 13, 2026
About this model
Description below is the original text supplied by OpenRouter, shown in the source language.
Drex v1.5 is a 9B decision model from Nace.AI. Given a state (text or JSON) and typed questions (yes/no, multiple choice, or ordinal score), it returns a calibrated probability for...
Specifications
- Model ID
- nace-ai/drex-v1.5
- Canonical ID
- nace-ai/drex-v1.5-20261009
- Creator
- nace-ai
- Added to OpenRouter
- 09.10.2026
- Context length
- 131,072
- Modality
- text->decisions
- Input modalities
- text
- Output modalities
- decisions
- Token pricing
- No per-token price in the source snapshot — see OpenRouter for the billing unit
- Knowledge cutoff
- —
- Expiration date
- —
Tasks and capabilities
A grade appears only where the source supplies a matching measurement for that task, the model is active and the cohort holds at least 10 comparable models. The capability column lists facts the source declares — supported formats, reasoning and tool parameters — which are not measured quality.
| Task | Grade | Source-declared capability |
|---|---|---|
| General quality | No measurement | — |
| Coding | No measurement | — |
| Agents | No measurement | — |
| Writing | No measurement | — |
| Reasoning | No measurement | — |
| Document analysis | No measurement | — |
| Vision | No measurement | — |
| Image generation | No measurement | — |
| Audio | No measurement | — |
| Video | No measurement | — |
| Embeddings | No measurement | — |
Evaluations
No evaluations are supplied for this model.
How grades work
Each letter grade compares a model only against other models scored by the same source and metric in the current snapshot. S means the top 10%, A the next 20%, B the next 30%, C the rest. Models without a published measurement are shown as not rated — a supported format alone never implies measured quality.
- S — S — top 10% of the cohort
- A — A — next 20% of the cohort
- B — B — next 30% of the cohort
- C — C — below the B band
Providers
Optional live provider statistics from OpenRouter. Missing values are unknown, not zero.
| Provider | Context length | Price in /1M | Price out /1M | Latency p50 (30m, s) | Throughput p50 (30m, tok/s) | Uptime (30m) |
|---|---|---|---|---|---|---|
| DeepInfra | 131,072 | — | — | — | — | 100% |
Provider data retrieved: 13.11.2026
Related articles
No published articles identify this model yet.